STRUCTURE is more robust than other clustering methods in simulated mixed-ploidy populations

Area of Science:

• Population Genetics
• Genomics
• Bioinformatics

Background:

• Population genetic structure analysis is crucial for understanding evolutionary processes.
• Clustering methods in polyploid complexes can reveal origins and admixture patterns.
• Combining diploid and polyploid data may introduce ploidy-based biases.

Purpose of the Study:

• To systematically compare clustering algorithms for analyzing mixed-ploidy genetic data.
• To evaluate method performance under varying population differentiation and marker types.
• To identify the most robust and efficient methods for inferring population structure.

Main Methods:

• Simulated mixed-ploidy (diploid-autotetraploid) datasets were generated.
• Performance of k-means, STRUCTURE, ADMIXTURE, FASTSTRUCTURE, and INSTRUCT was assessed.
• Scenarios included varying population differentiation and marker types (dominant, codominant).

Main Results:

• All tested methods performed well under strong population differentiation.
• STRUCTURE provided unbiased inference for markers with limited genotypic information when differentiation was weak.
• FASTSTRUCTURE is a faster alternative to STRUCTURE for large datasets, albeit less powerful.
• K-means clustering outperformed FASTSTRUCTURE in speed and power for large numbers of loci with known dosage.

Conclusions:

• STRUCTURE is the most robust method for analyzing genetic structure in mixed-ploidy populations.
• Alternative methods like FASTSTRUCTURE and k-means clustering can be suitable under specific conditions (dataset size, marker information).
• Careful consideration of method choice is essential for accurate inference in complex polyploid systems.